Method and apparatus for generating image textures

- Sarnoff Corporation

A method and apparatus for synthesizing texture images that match the texture appearance of a given digitized sample of an image. This synthesis is completely automatic and requires only the "target" texture image as an input. It allows generation of as much texture as desired so that any object or surface can be covered with the synthesized texture. The apparatus generates texture parameters for the target texture sample image that reflect the texture in that image. These texture parameters are used to modify the noise image such that the modified noise image attains the texture of the target texture image. To further improve the texture of the modified noise image, an image pyramid is constructed for both the modified noise image and the target texture sample image. The apparatus then iteratively updates the modified noise image on a coarse-to-fine basis using the subbands of the image pyramids. The output image is an updated modified noise image that has attained the texture of the target texture sample image.

Skip to:  ·  Claims  ·  References Cited  · Patent History  ·  Patent History

Claims

1. A method for generating image textures comprising the steps of:

(a) providing a first image having a texture;
(b) providing a second image;
(c) generating an image pyramid of said first image containing a first plurality of first subbands;
(d) generating an image pyramid of said second image containing a second plurality of second subbands, where said first subbands correspond to said second subbands;
(e) determining texture parameters for each of said subbands in said first and second pluralities of subbands;
(f) modifying, in response to the texture parameters of said first plurality of first subbands, said second subbands in said second plurality of second subbands until the texture parameters of said second subbands are substantially the same as the texture parameters of said first subbands to form a modified second image pyramid; and
(g) collapsing said modified second image pyramid to form a modified image having a texture that is substantially similar to the texture of said first image.

2. The method of claim 1 wherein said texture parameters comprise a histogram for each of said subbands in said first plurality of first subbands and said second plurality of second subbands.

3. The method of claim 1 wherein said (f) modifying step is performed in a coarse to fine subband order.

4. The method of claim 3 further comprising the step of:

(h) repeating steps (c), (d), (e), (f) and (g) using said modified image as said second image until a predefined criterion is met.

5. A method for generating image textures comprising the steps of:

(a) generating a histogram representing texture parameters of an input image;
(b) generating a histogram representing texture parameters of a noise image;
(c) modifying the noise image histogram to match the input image histogram, wherein said modified noise image attains the histogram of the input image;
(d) generating an image pyramid of said modified noise image containing a plurality of first subbands;
(e) generating an image pyramid of said input image containing a plurality of second subbands;
(f) generating, for each one of said plurality of first subbands, a histogram representing texture parameters of said each one of said plurality of first subbands;
(g) generating, for each one of said plurality of second subbands, a histogram representing texture parameters of said each one of said plurality of second subbands;
(h) selecting a subband from each of said plurality of first and second subbands;
(i) modifying said selected first subband histogram to match said selected second subband histogram, wherein said selected first subband histogram attains the histogram of said selected second subband histogram; and
(j) collapsing said modified image pyramid having a modified subband to form a second modified noise image.

6. The method of claim 5 wherein said (h) selecting step is repeated in a coarse to fine subband order and wherein said (i) modifying step is repeated for each selected subband.

7. The method of claim 6 further comprising the step of:

(k) repeating steps (d), (f), (h), (i) and (j) using said second modified noise image as said modified noise image until a predefined criterion is met.

8. The method of claim 1 further comprising the steps of:

providing a noise image;
generating a histogram of said noise image;
generating a histogram of said first image;
modifying said noise image such that said histogram of said noise image matches the histogram of said first image to form said second image.

9. The method of claim 8 wherein said noise image contains either random or pseudo-random noise.

10. The method of claim 1 wherein said second image is a noise image.

11. The method of claim 10 wherein said noise image contains either random or pseudo-random noise.

12. Apparatus for generating image textures having as inputs a first having a texture and a second image, said apparatus comprising:

an image pyramid generator for generating a first image pyramid from said first image having a first plurality of first subbands and a second image pyramid from said second image having a second plurality of second subbands, where said first subbands correspond to said second subbands;
a texture parameter generator, coupled to said image pyramid generator, for producing texture parameters for each of said first and second subbands;
image modifying means, coupled to said texture parameter generator, for altering said second subbands until said texture parameters of each said second subbands are substantially equal to said texture parameters of each of said first subbands to form a modified second pyramid; and
a pyramid processor, coupled to said texture parameter generator, for collapsing said modified second pyramid to form a modified second image having the texture of said first image.

13. The apparatus of claim 12 wherein said second image is a noise image.

14. The apparatus of claim 13 wherein said noise image contains either random or pseudo-random noise.

15. The apparatus of claim 12 further comprising means for generating said second image from a noise image including:

means for generating a histogram of said noise image;
means for generating a histogram of said first image;
means for modifying said noise image such that said histogram of said noise image matches the histogram of said first image to form said second image.

16. The apparatus of claim 15 wherein said noise image contains either random or pseudo-random noise.

17. The apparatus of claim 12 wherein said image modifying means operates upon said subbands in a coarse to fine subband order.

Referenced Cited
U.S. Patent Documents
5148497 September 15, 1992 Pentland et al.
5544292 August 6, 1996 Winser
Other references
  • Cadzow, James, "Image Texture Synthesis-by-Analysis Using Moving-Average Models", Oct. 1993 pp. 1110-1122. International Search Report mailed Feb. 13, 1997 for corresponding application No. PCT/US/96/12714. Lewis, "Texture Synthesis for Digital Painting", Computer Graphics, pp. 245-251, vol. 18, No. 3, Jul. 1984. Perlin, "An Image Synthesizer", Proceedings of SIGGRAPH '85, pp. 333-342, vol. 19, No. 3, Jul. 22-26, 1985. Ogden et al., "Pyramid-based Computer Graphics", RCA Engineer, 30-5, Sep./Oct./ 1985. Popat et al., "Novel Cluster-based Probability Model For Texture Synthesis, Classification, and Compression", Proc. of SPIE Visual Communications and Image Processing '93, Boston, Nov. 8-11, 1993. "Image Texture Synthesis-by-Analysis Using Moving-Average Models", James A. Cadzow, IEEE, vol. 29, No. 4 Oct. 1993.
Patent History
Patent number: 5872867
Type: Grant
Filed: Aug 4, 1995
Date of Patent: Feb 16, 1999
Assignee: Sarnoff Corporation (Princeton, NJ)
Inventor: James R. Bergen (Hopewell, NJ)
Primary Examiner: Leo H. Boudreau
Assistant Examiner: Daniel G. Mariam
Attorney: William J. Burke
Application Number: 8/511,258